“The Least Surprising Correlation Of All Time”

Greg Mankiw, Harvard economist, finds the inference often made (such as by the New York Times Economix Blog) from the correlation shown on a graph between parental income and SAT scores “The Least Surprising Correlation of All Time.” It tells us nothing, he says, “about the causal impact of income on test scores.”

The faulty inference, he writes, is a good example of “omitted variable bias.”

The key omitted variable here is parents’ IQ. Smart parents make more money and pass those good genes on to their offspring.

Suppose we were to graph average SAT scores by the number of bathrooms a student has in his or her family home. That curve would also likely slope upward. (After all, people with more money buy larger homes with more bathrooms.) But it would be a mistake to conclude that installing an extra toilet raises yours kids’ SAT scores.

It would be interesting to see the above graph reproduced for adopted children only. I bet that the curve would be a lot flatter.

There is extensive data on the other side of that coin, identical twins raised in different families, and even Nicholas Kristof begrudgingly recognizes that

a series of studies seemed to indicate that I.Q. is largely inherited. Identical twins raised apart, for example, have I.Q.’s that are remarkably similar. They are even closer on average than those of fraternal twins who grow up together.

No doubt there is also data showing that the SAT scores of twins raised apart are remarkably similar, but I don’t have time to look for it now.

UPDATE [30 August]

Reader Linda Seebach points to a fascinating follow-up to several criticisms of the Greg Mankiw post discussed above by Alex Tabarrok at Marginal Revolution. Those criticisms, by Paul Krugman and Matt Yglesias, and this one by Brad DeLong, all seek to minimize the influence of IQ, pointing to additional evidence showing strong correlations of college success and parental income.

“All this is good,” Tabarrok argues,

but none if it gets at the heart of the matter because there are a lot of way that heredity/genes could explain the income/education correlation; IQ is only one possible mechanism, personality (e.g. conscientiousness) is another possibility.

The type of evidence that we need to resolve this question is adoption studies. Fortunately, such studies have been done and indeed I have presented the data before in my post Nature, Nurture and Income. Let’s do so again.

Which he proceeds to do, with great effect. (But read the comments of many who are not persuaded.)

Say What? (2)

  1. David August 29, 2009 at 4:35 pm | | Reply

    I think Pat Moynihan was probably closest to the truth when he noted that states could move closer to Canada to raise their levels of educational achievement!

  2. M Stein August 30, 2009 at 4:50 am | | Reply

    Intelligence, like other behavioural traits is significantly heritable.

    “Data from more than 8000 parent-offspring pairs, 25,000 sibling pairs, 10,000 twin pairs and adoption studies provide evidence that genetic factors play a substantial role in the variation of general intelligence, with heritability estimates ranging from 40 to 80%” —Burdick et al, Cognitive variation in DTNBP1 influence general cognitive ability. Human Molecular Genetics, 2006, Vol 15, No. 10.

    “Heritability estimtes for intelligence quotient (IQ) range from 0.50 to 0.80. This makes IQ a suitable target for attempts to identify the specific genes involved.” Chorney et al, Role of the cholinergic muscarinic 2 receptor (CHRM@) gene in cognition. Molecular Psychiatry (2003) 8. 10-13.

    “A substantial body of literature from twin, family and adoption studies documents significant genetic effects on human intelligence. Heritability estimates range from 40 to 80% and meta-analyses suggest an overall heritability of around 50%” Dick et al,

    (2006) “Association of CHRM2 with IQ: Converging Evidence for Genes Influencing Intelligence.” Behavioral Genetics.

    “Multivariate genetic analyses indicate that general intelligence is highly heritable, and that the overlap in the cognitive processes is twice as great as the overall phenotypic overlap, with genetic correlations averaging around .80.”

    Plomin et al (2004) “A functional polymorphism in the succinate-semialdehyde dehydrogenase genes is associated with cognitive ability,” Molecular Psychology 9, 582-586.

    Also, the brains of more intelligent people are different in terms of cortical thickness and myelination (affects processing speed – Einstein had a larger number of glial cells which produces this). This is significantly heredible:

    “The UCLA researchers took the study a step further by comparing the white matter architecture of identical twins, who share almost all their DNA, and fraternal twins, who share only half. Results showed that the quality of the white matter is highly genetically determined, although the influence of genetics varies by brain area. According to the findings, about 85 percent of the variation in white matter in the parietal lobe, which is involved in mathematics, logic, and visual-spatial skills, can be attributed to genetics. But only about 45 percent of the variation in the temporal lobe, which plays a central role in learning and memory, appears to be inherited.”

    (see New Scientist 9 March 2009 ‘High Speed Brains in the Genes’ and MIT Technology Review, March 24, 2009)

Say What?